Heterogeneity of changes in separation rates

The last section documented that the decline in separation rates was the main driver of the reduction in unemployment rates in Germany after 2005. The decline of aver-age rates hides a lot of heterogeneity that is informative about the underlying causal mechanism. Figure 2.3 shows unemployment benefit eligibility by employment duration and age before and after the reform. This benefit duration determines when workers lose eligibility for UI benefits and transit to unemployment assistance benefits before and benefits at subsistence level after the reform as unemployment assistance benefits were abolished by the Hartz reforms. If this abolition of the unemployment assistance benefits is the driver of the observed changes in separation rates, we should see hetero-geneity in the changes of separation rates by employment duration and age. Looking at the pre-reform situation in Figure 2.3(a), we see that for workers younger than 45 the maximum benefit duration was 12 months. For older workers, we find a steep gradient in employment duration from 14 months after 30 months of previous employment to up to 30 months after five years of previous employment. Comparing this pattern to the post-reform regulation in Figure 2.3(b), we see that there is much less variation and that especially older, long-term employed workers see a strong decline in their benefit duration. For example, a 49-year-old worker with four years of previous employment receives, after the reform, UI benefits for 12 months, while before the reform she re-ceived UI benefits for 22 months. Figure 2.3(c) shows the relative changes in UI benefit durations for the different groups from before to after the reform. We find the largest decline for workers with more than three years of previous employment duration be-tween ages 45 and 55. By contrast, there have been no changes for short-term employed workers (less than 28 months) and workers younger than 45. If the Hartz reforms are causal for the decline in separation rates, we expect this heterogeneity in the changes in the duration of benefit eligibility to be mirrored in the changes in separation rates.

A further dimension where we should see differences in separation rate changes, if the causal mechanism is related to the abolition of the long-term benefits, is wages because unemployment assistance was tied to a worker’s last wage. A decoupling of long-term benefits from previous wages disproportionately affects workers with high wages because, after the reform, these workers face benefits at a subsistence level, independent of the previous wage, once UI benefits have expired.

We explore the changes in the separation rates along these dimensions of hetero-geneity. In line with a causal mechanism that works through the cut in long-term wage-dependent benefits, we find that long-term employed and high-wage workers show stronger declines in separation rates compared to short-term employed and low-wage workers.

Figure 2.3: Changes in benefit duration by age and employment duration

(a) Benefit duration pre-reform (b) Benefit duration post-reform

(c) Change in benefit duration

Notes: Maximum eligibility duration for short-term unemployment benefits in months by age and employment duration. Employment duration refers to a reference period of 5-7 years prior to the unemployment spell. Panel (a) shows maximum durations before the reform. Panel (b) shows the maximum durations after the reform in 2008, i.e. after all grandfathering rules had expired. Panel (c) shows the relative change in maximum durations in percent for each combination of age and employment duration.

Employment duration

For the analysis of heterogeneity among workers with different employment duration, we split employed workers into two groups. The first group is short-term employed workers with at most three years of employment duration, and the second group is long-term employed workers with more than three years of employment duration. Table 2.3 shows the corresponding average levels for the pre- and post-reform period. Looking at the levels, we see that short-term employed workers have separation rates that are more than five times higher than those of the long-term employed workers in the period 1993 to 2002 (1.37% vs 0.26%). This difference further increases in the period 2008 to 2014 (1.15% vs 0.18%). After 2008, separation rates differ by more than a factor of six.

The reason is the much stronger relative decline in the separation rate for long-term employed workers after 2008.

The stronger decline can be seen in Figure 2.4, which shows the time series of relative

Table 2.3: Change in separation rates by employment duration and age 1993 - 2002 2008 - 2014 ∆ %

all 0.63% 0.49% -22.0%

emp. duration ≤ 3 years 1.37% 1.15% -16.2%

emp. duration > 3 years 0.26% 0.18% -33.3%

Notes: Separation rates before and after the Hartz reforms by employment duration and age. We use averages of quarterly rates over the time periods. Column ∆ reports the percentage change in rates from the period before the Hartz reforms to the period after the Hartz reforms.

changes in separation rates for different groups of short-term and long-term employed workers. Looking at Figure 2.4(a), we find a strong divergence in the time series of separation rates between short-term and long-term employed workers after the Hartz reforms. The strong divergence persists so that, after the reform, separation rates of long-term employed workers have declined twice as much as those for short-term employed workers.

Age

In addition to employment duration, age determines the duration of benefit eligibility in the UI system (Figure 2.3). We therefore dissect the data in Figures 2.4(b) and 2.4(c) further by looking at young and old workers by employment duration. Looking at younger workers in Figures 2.4(b), we find, in line with the changes in eligibility duration, no differential changes between short-term and long-term employed workers.

Separation rates decline in lockstep for these two groups. By contrast, but in close alignment with the changes in eligibility duration, we find that long-term employed, older workers show the strongest reduction in separation rates, while older short-term employed workers show a reduction that is only half as large (Figure 2.4(c)). Look-ing at short-term employed workers across age groups provides a further sanity check for heterogeneous changes in separation rates because there have been no differential changes in benefit eligibility duration for short-term employed workers. In line with no such heterogeneity, we find a strikingly close tracking of separation rate changes for short-term employed young (age 15-44) and old workers (age 45-64) in Figure 2.4(d).

For both age groups separation rates decline in lockstep following the Hartz reforms.

These results by age further strengthen our finding that separation rates decline more for workers who have been more adversely affected by the cut in benefit eligibility from the Hartz reforms. Therefore, this additional heterogeneity in changes in separation rates provides further support for a causal link from the UI reform to the observed changes in labor market dynamics.

In Appendix C, we provide a more detailed analysis of changes by age groups. One finding from this analysis is that workers closer to retirement show an even stronger decline in separation rates. Their decline in separation rates follows a longer-run trend

Figure 2.4: Separation rates by age and employment duration (1993 - 2014)

(a) all workers

1995 2000 2005 2010 40

60 80 100 120 140 160

long-term short-term

(b) age 15 - 44

1995 2000 2005 2010 80

100 120

140 ^long-term_short-term

(c) age 45 - 64

1995 2000 2005 2010 40

60 80 100 120 140

160 long-term

short-term

(d) short-term employed

1995 2000 2005 2010 80

100 120 140

45-64 15-44

Notes: Separation rates by employment duration and age for West Germany 1993 - 2014, indexed to their pre-reform level (1993-2002). The red solid lines in panels (a)-(c) mark the separation rate for long-term employed workers who were continuously employed for three years or more. The blue dashed lines in panels (a)-(c) mark the separation rate for short-term employed workers with at most three years of continuous employment. Panel (d) shows the separation rate for short-term employed workers separately for young (blue dashed line) and old employees (red solid line). The grey area indicates the period of the implementation of the Hartz reforms. The grey area marks the period 2003 to 2005 when the Hartz reforms were enacted. The fading out indicates the first transition years 2006 to 2008 after the reforms. Data are quarterly averages of monthly rates.

that accelerated during the 2000s so that, over time, unemployment rates for older workers decreased more than those for younger workers. This trend was accompanied by a strongly rising labor force participation rate of workers close to retirement age. We abstract from this fact of independent interest as it is beyond the scope of this paper.¹²

12Jäger et al. (2018) provide a detailed investigation of this topic. They study changes in separation rates of male workers towards the end of working life (50 years and older) in Austria after changes in UI benefit durations. They exploit staggered changes in UI eligibility similar to those shown in

In our theoretical analysis, we will abstract from the additional age heterogeneity to keep the model parsimonious and because most of the heterogeneity in the changes in separation rates is captured by differences in employment duration.

Wages

Figure 2.5 shows the relative changes in separation rates from before to after the UI reform along the wage distribution. In a first step, we consider two groups of workers:

low-wage workers in the bottom three deciles of the wage distribution and high-wage workers in the fourth to seventh deciles. In a second step, we provide more granular changes showing that the differences further up in the wage distribution are, if anything, larger. Figure 2.5(a) compares changes in separation rates for low-wage workers and high-wage workers over time. Separation rates before 2005 comove closely and start to diverge thereafter. By 2014, high-wage workers saw their separation rates decline by almost twice as much as those for low-wage workers. Figure 2.5(b) dissects the wage distribution finer. It quantifies for each wage decile by how much the average separation rate decreased from the decade before the reform (1993-2002) to the post-reform period (2008-2014). Evidently, the higher wage deciles experienced the largest declines in separation rates. Separation rates hardly change at the bottom of the wage distribution, decline by between 20% to 30% in the middle, and plummet by almost 50% at the top. The stronger decline in separation rates for high-wage workers further supports a causal link from the UI reform to the observed changes in labor market dynamics because the reform replaced long-term wage-dependent benefits by subsistence benefits independent of previous wages. This change affected in particular high-wage workers.

The results on employment duration, age, and wage heterogeneity all show a larger drop in separation rates for groups that have been more adversely affected by the Hartz reforms (Figure 2.3). We speak to the observed heterogeneity in our quantitative model below. In the model, the removal of long-term wage-dependent benefits will lead to heterogeneous reactions in separation rates, and high-wage, long-term employed workers will see a stronger reduction in their separation rates in line with the empirical evidence from this section.